Dynamic Risk Trajectories for Sudden Cardiac Arrest: The Role of Recurrent Cardiovascular Events

Abstract

Background

Although cardiovascular events increase sudden cardiac arrest (SCA) risk, the impact of recurrent events on subsequent SCA risk in a contemporary population is unknown. This study assessed whether patients with a first‐time acute coronary syndrome (ACS) or heart failure (HF) hospitalization who experience a recurrent cardiovascular event have increased SCA risk.

Methods

The OSCAR (Observational Study of Cardiac Arrest Risk) is a prospective cohort with adjudicated SCA outcomes. The current study followed up patients who survived a first ACS or HF hospitalization (the index ACS or HF cohorts) for recurrent cardiovascular events and SCA. Recurrent event was a time‐dependent variable in Cox models predicting SCA. Findings were validated in the FHS (Framingham Heart Study).

Results

Among 2946 patients in the index ACS cohort, incidence of SCA was higher following a recurrent ACS event than without (3.70 versus 1.28 per 100 patient‐years). A recurrent ACS event was associated with higher risk of SCA (adjusted hazard ratio [HR], 3.15 [95% CI, 2.06–4.83]; P<0.0001). Among 6711 patients in the index HF cohort, incidence of SCA was higher following a recurrent HF event than without (1.35 versus 0.97 per 100 patient‐years), and SCA risk was higher with recurrent HF (HR, 1.79 [95% CI, 1.44–2.23]; P<0.0001). In the FHS cohort, risk of SCA was higher with recurrent ACS (HR, 2.85 [95% CI, 1.66–4.90]; P=0.0002); the association was not significant for recurrent HF (HR, 1.49 [95% CI, 0.73–3.03]; P=0.27).

Conclusions

Recurrent events were associated with higher risk of SCA; dynamic clinical trajectories of recurrent cardiovascular events may inform prevention of SCA.

Nonstandard Abbreviations and Acronyms

CIF

cumulative incidence function

FHS

Framingham Heart Study

HFpEF

heart failure with preserved ejection fraction

OSCAR

Observational Study of Cardiac Arrest Risk

SCA

sudden cardiac arrest

SCD

sudden cardiac death

Clinical Perspective.

What Is New?

• Among patients with a first acute coronary syndrome or first heart failure hospitalization, those with a recurrent acute coronary syndrome event or subsequent heart failure hospitalization, respectively, had increased risk of sudden cardiac arrest compared with those without recurrence.

What Are the Clinical Implications?

• These findings indicate that individuals with recurrent cardiovascular events are at higher risk of sudden cardiac arrest.

• Dynamic risk trajectories may inform patient management and improve prevention of sudden cardiac arrest, but prospective studies are needed to test these findings in larger and more diverse populations.

Sudden cardiac arrest (SCA) is a major public health challenge, resulting in substantial loss of life each year. In the United States alone, >350 000 individuals experience SCA annually, with survival rates remaining <10%. ¹ Current SCA risk prediction primarily relies on static measures, with left ventricular ejection fraction (LVEF) <35% serving as the principal selection criterion for primary preventive implantable cardioverter‐defibrillator implantation. ² , ³ However, this approach has limited specificity and sensitivity; advances in heart failure (HF) management have led to a reduction in SCA rates among patients with reduced LVEF, ⁴ and most (>70%) of patients who experience SCA in the community have LVEF of >35%. ⁵ Consequently, there is a critical need for improved risk stratification strategies. Emerging evidence suggests that dynamic risk assessment, accounting for changes in vulnerable clinical substrate over time, may improve SCA risk stratification. ⁶ , ⁷

Acute coronary syndrome (ACS) and HF are well‐established pathophysiological conditions that predispose to SCA. ⁸ Patients who experience recurrent cardiovascular events, including new ACS episodes or HF hospitalizations, are at elevated risk of major adverse cardiac events. ⁹ , ¹⁰ , ¹¹ , ¹² Recurrent cardiovascular events may reflect progression of underlying myocardial injury, ischemia, and electrical instability, or worsening ventricular remodeling, all of which can contribute to increased susceptibility to malignant arrhythmias. However, the association between recurrent cardiovascular events and the risk of SCA remains insufficiently characterized. A few post hoc analyses from randomized controlled trials (RCTs) have suggested an increased risk of sudden death among patients with ACS who experience recurrent cardiovascular events. ¹³ , ¹⁴ , ¹⁵ In patients with HF, some studies have reported a higher risk of malignant arrhythmias in those with recurrent events; however, these studies were restricted to implantable cardioverter‐defibrillator recipients and did not specifically evaluate SCA as an outcome. ¹⁶ , ¹⁷

In the current study, data from the large, prospective OSCER (Observational Study of Cardiovascular Event Risk) cohort were used to investigate the associated risk of SCA in patients with a first ACS or HF hospitalization who experienced a recurrent cardiovascular event compared with those who did not. The findings were independently validated in the FHS (Framingham Heart Study).

METHODS

OSCAR Cohort

Study Design

The OSCAR has been described in detail previously. ¹⁸ In brief, the OSCER cohort is an electronic health record (EHR) study of Los Angeles County residents with at least 1 patient encounter with the Cedars‐Sinai Health System in 2 consecutive years during the period from 2016 to 2020. The OSCAR consists of the subset with no prior SCA. EHR data were obtained from all outpatient and inpatient encounters across the health care levels from year 2012.

For the current study, we included patients aged ≥35 years with a first hospitalization for ACS or HF after enrollment in OSCAR in 2 distinct cohorts: an index ACS cohort and an index HF hospitalization cohort (Figure 1A). We only included patients who survived to hospital discharge from their index event to ensure a population truly at risk of experiencing a recurrent event. We excluded patients with a history of the same cardiovascular event type from the given cohort to avoid immortal time bias, as inclusion would require survival from the initial ACS to entry in OSCAR, a period during which death could not have occurred. Patients could contribute to both cohorts. We then prospectively followed up patients for recurrent event of the same index event type, and for the outcome of SCA. Only the first occurrence of a recurrent event was considered. Time at risk ended at time of SCA, death, or at the end of follow‐up. There were no criteria for survival time after a recurrent event. At the time of data extraction, follow‐up data were available until September 2024, making the maximum follow‐up time 7.75 years.

Figure 1. Illustration of the study design.

A, Identification of patients from the OSCAR to include in the ACS and HF hospitalization cohorts. Enrollment in OSCAR required at least 1 patient encounter within the Cedars‐Sinai Health System in 2 consecutive years. B, Identification of patients from the FHS to include in the ACS and HF hospitalization cohorts. ACS indicates acute coronary syndrome; FHS, Framingham Heart Study; HF, heart failure; OSCAR, Observational Study of Cardiac Arrest Risk; SCA, sudden cardiac arrest; and SCD, sudden cardiac death.

Variables and Clinical Event Definitions

We retrieved demographics, risk behavior, anthropometrics, and clinical comorbidities from the EHR during the baseline period. Clinical events and treatment during follow‐up were defined by appropriate International Classification of Diseases, Tenth Revision (ICD‐10 ), diagnostic and procedure codes as well as Current Procedural Terminology codes, and an ACS was defined as a hospitalization for unstable angina or myocardial infarction (MI), and HF hospitalization was a hospitalization where the primary ICD‐10 code was HF (ICD‐10 code list in Tables S1 and S2).

We identified cases of presumed SCA from the regional emergency medical services database. Thereafter, trained research physicians conducted a careful case adjudication process using a standardized method as well as all available data, including medical records, emergency medical services reports, and death certificates. ¹⁸ SCA was defined as a sudden and unexpected loss of pulse attributable to a likely cardiac cause, ¹⁹ and patients were excluded if the SCA occurred because of unnatural causes (trauma, suicide, or drug overdose) or natural noncardiac causes (eg, acute cerebrovascular accident; sepsis or pneumonia; or terminal illness, including metastatic cancer, Alzheimer disease, or other terminal neurologic diseases, such as multiple sclerosis/amyotrophic lateral sclerosis). Survivors of SCA (n=26, 5.1% of total SCAs) were included in the SCA outcome definition.

FHS Cohort

Study Design

The FHS is a community‐based epidemiologic study conducted in Framingham, Massachusetts. The study method has been described in detail previously. ²⁰ , ²¹ The study was initiated in 1948 with an original cohort of 5209 participants and was extended with an offspring cohort of 5124 participants in 1971. All participants underwent an initial interview and physical examination, followed by follow‐up examinations approximately every 2 to 4 years.

For the current study, we categorized participants with an index ACS or HF hospitalization during follow‐up into 2 corresponding cohorts, as done in the OSCAR cohort (Figure 1B). Included participants were aged ≥35 years at the time of the index event. As data on hospital length of stay were not available in FHS, survival to hospital discharge was approximated by including only patients who survived 14 days after the index event. We followed up patients for recurrent cardiovascular event and the sudden cardiac death (SCD) outcome, with maximum follow‐up time set at 8 years as the maximum follow‐up time in OSCAR was 7.75 years.

Variables and Clinical Event Definitions

Comorbidities were defined by the FHS clinical criteria. Baseline characteristics were taken from the latest clinical examination before the qualifying index event, and medical history was compiled from all examinations up to the most recent one before the index event.

Cardiovascular events in FHS participants were reviewed and adjudicated by a committee of 3 physicians. ACS was defined as presence of at least 2 of the following 3 criteria: ischemic symptoms, changes in biomarkers reflecting myocardial necrosis, and dynamic ECG changes indicating myocardial ischemia, reflecting the diagnostic criteria of ACS in that period. HF diagnosis was based on the Framingham HF criteria, with presence of at least 2 major or 1 major and 2 minor criteria for HF. SCD was defined as deaths occurring within 1 hour of symptom onset, with no other probable cause of death other than coronary heart disease. ²² In the FHS cohorts, only fatal SCA cases were included, as this was the outcome available.

Statistical Analysis

We report continuous variables as means (SDs) and median (interquartile range [IQR]) and categorical variables as frequencies and percentages.

Incidence rates are expressed as events per 100 patient‐years and calculated by dividing the number of events by the time at risk. For overall incidence rate of SCA, time at risk was taken from the index event to SCA, other mortality, or end of follow‐up. When calculating the specific incidence rate of SCA in patients with and without a recurrent cardiovascular event, time at risk was defined as follows: for patients without a recurrent event, time at risk extended from the index event to SCA, other mortality, or end of follow‐up; for patients with a recurrent event, time at risk started at the time of the first recurrent event and ended with SCA, other mortality, or end of follow‐up. The period between the index event and the first recurrence was considered unexposed time and included in the nonexposed group, thereby avoiding misclassification or exclusion of immortal time, which can bias results. ²³ , ²⁴

We used cumulative incidence functions (CIFs) to estimate the cumulative incidence, as traditional Kaplan–Meier curves can overestimate risk in the presence of competing mortality risk. ²⁵

To assess the risk associated with a recurrent cardiovascular event during follow‐up, we used cause‐specific Cox proportional hazards models with the recurrent cardiovascular event treated as a time‐dependent covariate. ²³ In the adjusted Cox model, covariates included were demographics, cardiovascular risk factors, and comorbidities, and a backward stepwise selection procedure was applied, retaining variables with P<0.20. In the OSCAR cohorts, the following variables were tested: age, sex, race and ethnicity, history of smoking, body mass index, hypercholesterolemia, hypertension, diabetes, chronic obstructive pulmonary disease, chronic kidney disease, peripheral arterial disease, coronary artery disease, HF (in the ACS cohort), type of index ACS (ST‐segment–elevation MI, non–ST‐segment–elevation MI, or unstable angina) and revascularization for index ACS event (in ACS cohort), and history of ACS (in HF cohort). In the Framingham cohorts, some variables included in the OSCAR models were not available; others were available but had a large proportion of missing data. These variables were not included in FHS models. Variables available and tested were age, sex, history of smoking, body mass index, hypercholesterolemia, hypertension, diabetes, HF (in ACS cohort), history of ACS (in HF cohort), and year of index cardiovascular event. In OSCAR models, missing data (<2% for any variable) were recoded as absence of the condition. In FHS data, missing data are noted in table footnotes. To retain individuals with missing hyperlipidemia data (12%) in multivariable models, we created a “missing” category. Data for smoking and hypertension were missing for <3.6% of the FHS data set and recoded as absence of each condition.

For all Cox models, age was treated as a time‐dependent variable. To evaluate whether inclusion of the recurrent event variable significantly improved the prediction of SCA outcomes, we calculated the Harrell concordance index (Harrell C) using the “concordance” function (R survival package) for models with and without the recurrent event variable in the ACS and HF cohorts, respectively. We computed the P value using the “compare_C_stat” function in R. We evaluated the proportional hazards assumption using Schoenfeld residuals; the global P value for the OSCAR ACS cohort was 0.75, and the value for the HF cohort was 0.77, suggesting no violation of the proportional hazards assumption.

As a sensitivity analysis, we used Fine‐Gray subdistribution hazard ratios (HRs) to account for competing risks, treating non‐SCA mortality as a competing event.

We also performed a sensitivity analysis stratifying the OSCAR HF population by LVEF into HF with preserved ejection fraction (LVEF ≥50%) and HF with reduced ejection fraction (HFrEF; LVEF ≤40%) to assess whether the association between recurrent HF hospitalization and SCA risk differed by systolic function.

All statistical analyses were performed using SAS version 9.4 (SAS Institute Inc, Cary, NC). Statistical significance was defined as a 2‐sided P≤0.05.

Ethical Approval

The OSCAR was approved by the institutional review board at Cedars‐Sinai Health System (CSHS) and by the Honest Enterprise Research Broker Committee of CSHS's Enterprise Information Systems Research Informatics and Scientific Computing Core.

The FHS was approved by the institutional review board of Boston University Medical Center, and all participants provided written informed consent. The current research was conducted as an ancillary study approved by the FHS Executive Committee.

The study data will be made available by the corresponding author on reasonable request.

RESULTS

OSCAR Cohort

ACS Cohort: Patient Population

The OSCAR ACS cohort comprised 2946 patients, and their baseline clinical characteristics overall and according to outcome can be found in Table 1. Patients who experienced SCA had a higher prevalence of several comorbidities, including diabetes, chronic kidney disease, and peripheral arterial disease, compared with those who did not experience SCA. The proportion of patients with a recurrent ACS event was highest among those who subsequently experienced SCA (SCA: 24.6%; other mortality: 14.9%; alive: 11.8%).

Table 1.

Clinical Characteristics According to Outcomes in the OSCAR Index Cohort With ACS

Clinical characteristics	Total (n=2946)	Alive (n=1731)	SCA (n=122)	Other mortality (n=1093)
Age, mean±SD, y	73.9±13.8	70.1±13.1	74.6±14.2	79.8±12.6
Male sex	1597 (54.2)	946 (54.7)	66 (54.1)	585 (53.5)
Race or ethnicity
Asian	214 (7.3)	131 (7.6)	14 (11.5)	69 (6.3)
Black	557 (18.9)	310 (17.9)	38 (31.2)	209 (19.1)
Hispanic	273 (9.3)	169 (9.8)	11 (9.0)	93 (8.5)
White	1789 (60.7)	1051 (60.7)	57 (46.7)	681 (62.3)
Other	113 (3.8)	70 (4.0)	2 (1.6)	41 (3.8)
Obesity	821 (27.9)	538 (31.1)	42 (34.4)	241 (22.0)
Hyperlipidemia	1903 (64.6)	1109 (64.1)	76 (62.3)	718 (65.7)
History of smoking	1078 (36.6)	603 (34.8)	57 (46.7)	418 (38.2)
History of HF	1790 (60.8)	832 (48.1)	96 (78.7)	862 (78.9)
Diabetes	1048 (35.6)	547 (31.6)	65 (53.3)	436 (39.9)
Hypertension	2146 (72.8)	1165 (67.3)	107 (87.7)	874 (80.0)
COPD	503 (17.1)	216 (12.5)	28 (23.0)	259 (23.7)
CKD	747 (25.4)	280 (16.2)	61 (50.0)	406 (37.1)
PAD	615 (20.9)	278 (16.1)	49 (40.2)	288 (26.3)
Known CAD (without ACS) before index event	1067 (36.2)	565 (32.6)	52 (42.6)	450 (41.2)
Index ACS event
NSTEMI	1905 (64.7)	987 (57.0)	90 (73.8)	828 (75.8)
STEMI	159 (5.4)	92 (5.3)	5 (4.1)	62 (5.7)
Unstable angina	387 (13.1)	334 (19.3)	6 (4.9)	47 (4.3)
Unspecified	495 (16.8)	318 (18.4)	21 (17.2)	156 (14.3)
Revascularization for index ACS	921 (31.3)	689 (39.8)	25 (20.5)	207 (18.9)

Data are given as number (percentage) unless otherwise indicated. "Other" includes American Indian/Alaskan Native, Native Hawaiian/Pacific Islander, and Unknown race. ACS indicates acute coronary syndrome; CAD, coronary artery disease; CKD, chronic kidney disease; COPD, chronic obstructive pulmonary disease; HF, heart failure; NSTEMI, non–ST‐segment–elevation myocardial infarction; OSCAR, Observational Study of Cardiac Arrest Risk; PAD, peripheral arterial disease; SCA, sudden cardiac arrest; and STEMI, ST‐segment–elevation myocardial infarction.

ACS Cohort: Incidence of SCA and Influence of Recurrent Event

Over a median follow‐up of 2.3 (IQR, 0.6–4.5) years, 1215 patients died. In total, 122 patients experienced SCA. The incidence rate of SCA was 1.53 (95% CI, 1.27–1.82) per 100 patient‐years. A recurrent ACS event occurred in 397 (13.5%) patients, and the incidence rate of SCA was greater following a recurrent event compared with no recurrent event (3.70 [95% CI, 2.50–5.29] versus 1.28 [95% CI, 1.03–1.57]). The CIF for SCA after a recurrent event and in those without a recurrent event can be seen in Figure 2A. At 5 years of follow‐up, the CIF of SCA after a recurrent event was 9.6% (95% CI, 6.5%–13.4%), and without a recurrent event, it was 3.9% (95% CI, 3.1%–4.8%).

Figure 2. Cumulative incidence function of SCA in patients after a recurrent event (red) and without a recurrent event (blue) in the OSCAR ACS and HF cohorts.

A, In the OSCAR ACS cohort, cumulative incidence of SCA was significantly higher (Gray test P<0.0001) among individuals with recurrent SCA than among individuals without recurrent ACS. The tables show outcomes and censoring over the first 5 years of follow‐up. All individuals in the ACS cohort entered at time 0 (on left, n=2946) and were followed up for the primary outcome of SCA or were censored. Time 0 (the ACS event) was between January 1, 2017, and September 30, 2024, with a median follow‐up of 2.3 years (minimum, 0.01 years; maximum, 7.75 years). Censoring from the ACS cohort (left) occurred if an individual had other cause mortality (competing event), a recurrent ACS event, or was alive but follow‐up time ended during that time interval. Individuals who experienced recurrent ACS stopped contributing time to the ACS cohort (were censored) and began contributing time to the recurrent ACS cohort (n=397 recurrent ACS [of these, 387 occurred within the first 5 years of follow‐up]). Individuals with recurrent ACS were followed up for the primary outcome of SCA or were censored (other cause mortality or alive but follow‐up time ended). B, In the OSCAR HF cohort, cumulative incidence of SCA was significantly higher (Gray P=0.0002) among individuals with recurrent HF hospitalization than among individuals without recurrent HF hospitalization. The tables show outcomes and censoring over the first 5 years of follow‐up. All individuals in the initial HF hospitalization cohort (n=6711) were followed up for the primary outcome of SCA or were censored. Entry into the cohort at time 0 (initial HF hospitalization) was between January 1, 2017, and September 30, 2024, with a median follow‐up 2.0 years (minimum, 0.01 years; maximum, 7.75 years). Censoring from the HF cohort occurred if an individual had other cause mortality (competing event), a recurrent HF hospitalization, or was alive but follow‐up time ended during that time interval. Individuals who experienced recurrent HF stopped contributing time to the HF cohort (were censored) and began contributing time to the recurrent HF hospitalization cohort (n=3806 recurrent HF hospitalizations [of these, 3767 occurred within the first 5 years of follow‐up]). Individuals with recurrent HF hospitalization were followed up for the primary outcome of SCA or were censored (other cause mortality or alive but follow‐up time ended). ACS indicates acute coronary syndrome; HF, heart failure; OSCAR, Observational Study of Cardiac Arrest Risk; and SCA, sudden cardiac arrest.

In the cause‐specific Cox proportional hazard model with recurrent event as a time‐dependent variable, the crude HR for SCA with a recurrent event was 3.51 (95% CI, 2.30–5.36; P<0.0001). After multivariable adjustment, the HR was 3.15 (95% CI, 2.06–4.83; P<0.0001). In the adjusted model, recurrent event was the variable with the highest associated risk of SCA (Table S3). The Harrell C‐index for the model, including the recurrent event with baseline covariates (C=0.756 [95% CI, 0.742–0.801]) was significantly higher than the model including baseline covariates only (C=0.743 [95% CI, 0.714–0.787]), P<0.001.

HF Cohort: Patient Population

The OSCAR HF cohort comprised 6711 patients, whose clinical characteristics both according to outcome and overall can be found in Table 2. Most patients had preserved LVEF (≥50; 64.8%). As in the ACS cohort, patients who experienced SCA had higher prevalence of several comorbidities, including diabetes, chronic kidney disease, and ACS, compared with those who did not experience SCA. A recurrent HF hospitalization occurred more often in patients who experienced SCA than in those who died of other causes (60.7% versus 53.3%).

Table 2.

Clinical Characteristics According to Outcomes in the OSCAR Index Cohort With HF

Clinical characteristics	All (n=6711)	Alive (n=3354)	SCA (n=387)	Other mortality (n=2970)
Age, mean±SD, y	77.2±13.3	73.6±13.0	77.0±13.8	81.2±12.3
Male sex	3524 (52.5)	1777 (53.0)	227 (58.7)	1520 (51.2)
Race or ethnicity
Asian	365 (5.4)	198 (5.9)	23 (5.9)	144 (4.85)
Black	1281 (19.1)	680 (20.3)	85 (22.0)	516 (17.37)
Hispanic	603 (9.0)	333 (9.9)	36 (9.3)	234 (7.9)
White	4223 (62.9)	2008 (59.9)	232 (60.0)	1983 (66.8)
Other	239 (3.6)	135 (4.0)	11 (2.8)	93 (3.1)
Obesity	2040 (30.4)	1174 (35.0)	122 (31.5)	744 (25.1)
Hyperlipidemia	4424 (65.9)	2142 (63.9)	264 (68.2)	2018 (67.9)
History of smoking	2519 (37.5)	1197 (35.7)	143 (37.0)	1179 (39.7)
HF	2146 (32.0)	852 (25.4)	168 (43.4)	1126 (37.9)
LVEF (%) at baseline, mean±SD*	52.7±16.3	52.4±16.3	49.7±16.7	53.4±16.1
≤40	1464 (24.9)	752 (25.5)	107 (31.6)	605 (23.4)
41–49	603 (10.3)	319 (10.8)	40 (11.8)	244 (9.4)
≥50	3802 (64.8)	1874 (63.6)	192 (56.6)	1736 (67.2)
Diabetes	2567 (38.3)	1190 (35.5)	187 (48.3)	1190 (40.1)
Hypertension	5361 (79.9)	2535 (75.6)	324 (83.7)	2502 (84.2)
COPD	1384 (20.6)	535 (16.0)	95 (24.5)	754 (25.4)
CKD	2027 (30.2)	763 (22.7)	167 (43.2)	1097 (36.9)
PAD	1644 (24.5)	689 (20.5)	129 (33.3)	826 (27.8)
CAD	3046 (45.4)	1337 (39.9)	198 (51.2)	1511 (50.9)
Prior ACS	2596 (38.7)	1239 (36.9)	183 (47.3)	1174 (39.5)

Data are given as number (percentage) unless otherwise indicated. "Other" includes American Indian/Alaskan Native, Native Hawaiian/Pacific Islander, and Unknown race. ACS indicates acute coronary syndrome; CAD, coronary artery disease; CKD, chronic kidney disease; COPD, chronic obstructive pulmonary disease; HF, heart failure; LVEF, left ventricular ejection fraction; OSCAR, Observational Study of Cardiac Arrest Risk; PAD, peripheral arterial disease; and SCA, sudden cardiac arrest.

^*Missing n=842 (12.6%).

HF Cohort: Incidence of SCA and Influence of Recurrent Event

Over a median follow‐up of 2.0 (IQR, 0.7–4.1) years, 3357 patients died. In total, 387 patients experienced an SCA. The incidence rate of SCA was 2.25 (95% CI, 2.03–2.49) per 100 patient‐years. A recurrent HF hospitalization occurred in 3806 (56.7%) patients, and the incidence rate of SCA was higher following a recurrent event compared with no recurrent event (2.63 [95% CI, 2.31–2.99] versus 1.84 [95% CI, 1.56–2.16]). The CIF for SCA after a recurrent event and without a recurrent event can be seen in Figure 2B. At 5 years of follow‐up, the CIF of SCA after a recurrent event was 8.1% (95% CI, 7.1%–9.3%), and without a recurrent event, it was 5.1% (95% CI, 4.2%–6.1%).

In the cause‐specific Cox proportional hazard model, the crude HR for SCA after a recurrent event was 1.82 (95% CI, 1.47–2.27; P<0.0001). After multivariable adjustment, the HR was 1.79 (95% CI, 1.44–2.23; P<0.0001) (Figure 3). In the adjusted model, recurrent event had the largest HR among all variables examined (Table S4). In the HF cohort, inclusion of recurrent HF with baseline covariates improved model performance (C=0.639 [95% CI, 0.606–0.684]) compared with the model with baseline covariates only (C=0.617 [95% CI, 0.604–0.646]), P<0.001.

Figure 3. Forest plot of the risk of SCA associated with a recurrent cardiovascular event among patients with an index cardiovascular event.

Risks were estimated using cause‐specific Cox proportional hazard models, with recurrent cardiovascular events modeled as a time‐dependent variable. *Variables included in the adjusted models: OSCAR ACS cohort: age, race and ethnicity, history of smoking, hypercholesterolemia, diabetes, CKD, PAD, HF, and revascularization for index ACS event; FHS ACS cohort: age, sex, diabetes, and year of index event; OSCAR HF cohort: age, sex, hypercholesterolemia, diabetes, CKD, PAD, prevalent HF, and history of ACS; FHS HF cohort: age, sex, and year of index event. ACS indicates acute coronary syndrome; CKD, chronic kidney disease; FHS, Framingham Heart Study; HF, heart failure; HT, hypertension; OSCAR, Observational Study of Cardiac Arrest Risk; PAD, peripheral arterial disease; and SCA, sudden cardiac arrest.

FHS Cohort

ACS Cohort

The FHS ACS cohort included 1631 patients. Compared with the OSCAR cohort, the FHS ACS cohort was slightly younger and had a lower comorbidity burden (Table 3).

Table 3.

Clinical Characteristics According to Outcomes in the FHS Index Cohort With ACS

Clinical characteristics	All (n=1631)	Alive (n=924)	SCD (n=79)	Other mortality (n=628)
Age, mean±SD, y	67.1±12.5	62.9±11.4	63.1±11.3	73.7±11.2
Male sex	1031 (63.2)	622 (67.3)	66 (83.5)	343 (54.6)
Obesity	362 (22.2)	215 (23.3)	14 (17.7)	133 (21.2)
Hyperlipidemia*	133 (9.3)	83 (10.3)	2 (3.4)	48 (8.5)
History of smoking†	602 (38.3)	401 (44.9)	32 (45.7)	169 (27.8)
HF	160 (9.8)	48 (5.2)	6 (6.3)	106 (16.9)
Diabetes	175 (10.7)	69 (7.5)	12 (15.2)	94 (15.0)
Hypertension‡	565 (34.7)	259 (28.1)	25 (32.1)	281 (44.7)

Data are given as number (percentage) unless otherwise indicated. ACS indicates acute coronary syndrome; FHS, Framingham Heart Study; HF, heart failure; and SCD, sudden cardiac death.

^*Missing n=200.

^† Missing n=58.

^‡ Missing n=3.

Over a median follow‐up time of 8.0 (IQR, 3.9–8.0) years, 707 patients died, of whom 11.2% (n=79) died of SCD. A total of 296 patients experienced a recurrent ACS event. The incidence rate of SCD following a recurrent event was higher than in those without a recurrent event (2.09 [95% CI, 1.24–3.30] versus 0.68 [95% CI, 0.52–0.87] per 100 patient‐years).

In the cause‐specific Cox proportional hazard model, the crude HR for SCD after a recurrent event was 3.12 (95% CI, 1.80–5.42; P<0.0001), and the adjusted HR was 2.85 (95% CI, 1.66–4.90; P=0.0002) (Figure 3; Table S5).

HF Cohort

In the FHS HF cohort, a total of 1081 patients were included. Baseline characteristics can be found in Table 4. Overall, the FHS HF cohort had a lower comorbidity burden compared with the OSCAR cohort.

Table 4.

Clinical Characteristics According to Outcomes in the FHS Index Cohort With HF

Clinical characteristics	All (n=1081)	Alive (n=260)	SCD (n=43)	Other mortality (n=778)
Age, mean±SD, y	77.2±11.3	72.2±11.5	73.9±10.7	79.0±10.7
Male sex	454 (42.0)	99 (38.1)	28 (65.1)	327 (42.0)
Obesity	325 (30.1)	106 (40.8)	13 (30.2)	206 (26.5)
Hyperlipidemia*	193 (19.0)	57 (23.1)	3 (7.9)	133 (18.2)
History of smoking†	201 (18.9)	44 (17.2)	13 (30.2)	144 (18.8)
ACS	159 (14.7)	48 (18.5)	5 (11.6)	106 (13.6)
Diabetes	167 (15.4)	29 (11.2)	7 (16.3)	131 (16.8)
Hypertension‡	585 (54.2)	132 (50.8)	15 (34.9)	438 (56.4)
LVEF (%) at baseline, mean±SD§	74±6	74±5	73±5	75±6
<50	2 (0.4)	0	0	2 (0.6)
≥50	504 (99.6)	135 (100)	19 (100)	349 (99.4)

Data are given as number (percentage) unless otherwise indicated. ACS indicates acute coronary syndrome; FHS, Framingham Heart Study; HF, heart failure; LVEF, left ventricular ejection fraction; and SCD, sudden cardiac death.

^*Missing n=65.

^† Missing n=16.

^‡ Missing n=1.

^§ Missing n=575.

Over a median follow‐up time of 3.2 (IQR, 0.7–7.8) years, 821 patients died, of whom 5.2% (n=42) died of SCD. In total 450 (56.7%) patients experienced a recurrent HF hospitalization. The incidence rate of SCD following a recurrent event was higher than without a recurrent event (1.35 [95% CI, 0.67–2.41] versus 0.97 [95% CI, 0.67–1.37] per 100 patient‐years).

In the cause‐specific Cox proportional hazard model, the crude HR for SCA after a recurrent HF hospitalization was 1.53 (95% CI, 0.77–3.05; P=0.236), and the adjusted HR was 1.49 (95% CI, 0.73–3.03; P=0.273) (Figure 3; Table S6).

Sensitivity Analyses

A competing‐risks sensitivity analysis was conducted in the OSCAR cohorts, in which we calculated the subdistribution HR for SCA treating other‐cause mortality as a competing risk event. In the ACS cohort, the subdistribution HR for SCA associated with a recurrent event was 3.13 (95% CI, 2.03–4.83; P<0.0001), and in the HF cohort, it was 2.02 (95% CI, 1.64–2.50; P<0.0001).

In a sensitivity analysis stratifying the OSCAR HF population by LVEF into HF with preserved ejection fraction and HFrEF, recurrent HF hospitalization was associated with significant increased risk of SCA in both groups (HF with preserved ejection fraction: adjusted HR, 1.49 [95% CI, 1.10–2.03], P=0.01; HFrEF: adjusted HR, 1.67 [95% CI, 1.08–2.58]; P=0.02).

In the OSCAR ACS cohort, a second recurrent ACS occurred in <3% of the cohort. In the OSCAR HF cohort, multiple HF hospitalization recurrences were common (Figure 2B). As a sensitivity analysis, we entered the number of HF hospitalizations in a competing risk model as an ordinal variable from 1 to 4. The HR increased from 1.47 for the first recurrent HF hospitalization to 5.01 for the fourth recurrent HF hospitalization. Cause‐specific HRs for other mortality also increased with the number of HF recurrences, but less steeply than SCA HRs.

DISCUSSION

In this observational prospective cohort study of patients with a first‐time ACS or HF hospitalization, recurrent cardiovascular events were associated with significantly higher risk of SCA. Specifically, in the index ACS cohort, a recurrent ACS event was associated with >3‐fold higher risk of SCA. Similarly, in the index HF cohort, a recurrent HF hospitalization was associated with a >80% higher risk of SCA. These findings were validated with similar levels of increased risk in the external cohorts from the FHS, although not reaching significance level for the HF cohort. To our knowledge, this is the first study to examine recurrent ACS or HF events as a potential dynamic risk marker for SCA in a contemporary, real‐world population.

Predicting SCA remains a significant challenge. Both ACS and HF are well‐established major risk factors for SCA, as they can lead to cardiac electrical and structural remodeling that increases susceptibility to malignant arrythmias. It is also recognized that SCA risk evolves over time, influenced by changes in the underlying substrate as well as transient triggers. ⁷ Among survivors of a first cardiovascular event, the risk of recurrent events, including new ACS episodes or HF hospitalizations, as well as cardiovascular and all‐cause mortality, is elevated. ⁹ , ¹⁰ , ¹¹ Moreover, mortality risk is further increased among those who do experience a recurrent event. ¹² However, the specific risk of SCA in this context has not been well characterized. In line with our results, a few post hoc analyses of RCTs have indicated a higher risk of SCA following recurrent ischemic events, ¹³ , ¹⁴ , ¹⁵ whereas the risk after recurrent HF hospitalizations has not been previously reported.

Recurrent ACS and Risk of SCA

In the OSCAR cohort of patients who experienced a first ACS, 13.5% went on to experience a new ACS event over a median follow‐up period of 2.3 years. This rate of recurrent events aligns with findings from another large, US‐based EHR medical claims cohort of a contemporary population with ACS. ¹¹

In both the ACS cohorts from OSCAR and FHS, the incidence rate of SCA was higher following a recurrent ACS event compared with those without a recurrent event. An EHR‐based study conducted in Finland on patients with ACS undergoing invasive evaluation similarly reported a higher cumulative incidence of SCA among patients who experienced a recurrent MI compared with those who did not (10.4% versus 6.2% at 12 years’ follow‐up). ¹⁵ Furthermore, in a post hoc analysis of the IMPROVE‐IT (Improved Reduction of Outcomes: Vytorin Efficacy International Trial), an RCT involving stabilized patients with ACS, the risk of sudden death was likewise higher after a recurrent cardiovascular event (composite of ACS, stroke, and hospitalization for HF) than in those without such an event (1.45 versus 0.27 per 100 patient‐years). ¹⁴

In the current study, survival analyses showed that a recurrent ACS event during follow‐up was associated with a 3‐fold higher risk of SCA in both the OSCAR and FHS cohorts. A sensitivity analysis, accounting for competing mortality risk, supported these findings with similar results. These observations are consistent with results from the IMPROVE‐IT trial, as well as in a pooled RCT analysis of patients with non–ST‐segment–elevation MI, reporting a 3 times higher risk of sudden deaths in patients who experienced a recurrent cardiovascular event. ¹³ Furthermore, in a post hoc analysis of the VALIANT (Valsartan In Acute myocardial Infarction trial) V RCT involving patients with MI and HF, landmark analyses for different time intervals showed that recurrent MI, HF hospitalization, or hospitalization for any cause were consistently associated with a significantly increased risk of SCD. ²⁶ Unlike these earlier published studies, the current study included all patients with a first‐time ACS event, specifically examined recurrent ACS events as the exposure (rather than a composite exposure variable), and used comprehensively adjudicated SCA outcomes. This enhances the clinical relevance and applicability of our findings. A prior study ²⁷ reported that recurrent ischemic events were not significantly associated with increased coronary artery–related SCD risk (HR, 1.26; P=0.09). However, the Adabag study was not contemporaneous (1979–2005) and was limited to Olmsted County, Minnesota. The current study was conducted in a multiethnic population under modern treatment patterns, making the findings more generalizable and relevant for today's clinical practice. The competing risk analysis performed in the current study further strengthens the robustness of the findings.

Recurrent HF and Risk of SCA

In the OSCAR cohort of patients experiencing a first HF hospitalization, 56.7% of patients experienced a recurrent HF hospitalization over a median follow‐up time of 2.0 years. Reported proportions in the literature vary depending on HF severity, phenotype, and comorbidity burden, but consistently indicate that recurrent hospitalizations are common. ²⁸ , ²⁹

Several studies have reported a significantly increased risk of overall mortality in patients with repeated HF hospitalizations. ²⁸ , ³⁰ , ³¹ , ³² , ³³ To our knowledge, no prior studies have specifically examined difference in SCA risk between patients who do or do not experience a recurrent HF hospitalization. Among patients with HF, some studies have suggested a higher risk of malignant arrhythmias in those with recurrent HF hospitalizations. A post hoc analysis of the implantable cardioverter‐defibrillator arm of the MADIT‐II (Multicenter Automatic Defibrillator Implantation Trial), including patients with MI and HFrEF, reported that HF hospitalization was associated with an HR of 2.5 for appropriate implantable cardioverter‐defibrillator shocks. Similarly, a pooled analysis of RCTs involving patients with both ischemic and nonischemic cardiomyopathy and HFrEF reported that HF hospitalization was associated with an HR of 1.79 for ventricular tachyarrhythmias. Although these studies were restricted to patients with HFrEF and focused on ventricular tachycardias as outcomes, they provide a similar signal to our current study. In the current study, recurrent HF hospitalizations in the OSCAR cohort were associated with a 79% increased risk of SCA, and this association remained consistent in a sensitivity analysis that accounted for other causes of mortality as a competing risk. Notably, the association between recurrent HF hospitalization and SCA risk was observed across the spectrum of systolic function. Although the associated increased risk was comparable in OSCAR and FHS, the finding was nonsignificant in the FHS, likely because of the smaller sample size and fewer SCD events.

Dynamic Risk Trajectories

Across both the index ACS and HF cohorts, the elevated risk of SCA following a recurrent cardiovascular event was consistent across Cox models (crude and adjusted) and in competing‐risk analyses, underscoring the robustness of the association, although in the FHS study, the association of increased SCA risk with recurrent HF was not statistically significant. These findings support a dynamic‐risk framework, wherein temporal changes in clinical characteristics during follow‐up provide incremental prognostic information beyond baseline factors. ⁷ The pattern aligns with emerging evidence that longitudinal trajectories, rather than a single time‐point measurement or static risk assessment, offer superior risk stratification. We have previously reported that ECG remodeling over time is itself a significant risk marker for SCA, reinforcing the value of dynamic assessment. ⁶ Future investigation should extend this approach by identifying additional dynamic risk markers and potentially developing risk‐prediction models that incorporate these factors to improve SCA prediction.

Strengths and Limitations

The OSCAR study enables ascertainment of a large, contemporary, real‐world cohort of patients with index ACS or index HF hospitalizations, and the unique integration of both EHR and emergency medical services data facilitates a robust and comprehensive adjudication process for SCA. Importantly, findings from OSCAR on recurrent ACS were replicated in the well‐established FHS, strengthening the validity and generalizability of the results. Estimated HRs for recurrent HF in the FHS were of similar magnitude to those in OSCAR, but not statistically significant.

Although both studies support the main findings, there are notable differences between the OSCAR and FHS cohorts. OSCAR is an EHR‐based cohort of patients who receive regular care within an advanced health care system, whereas FHS, in contrast, is a community‐based cohort designed to identify cardiovascular risk factors in included participants who were unselected for conditions at baseline. In addition, the FHS was initiated in an earlier era with different clinical practices and treatment options for ACS and HF. This could potentially explain why the OSCAR cohorts were older, had higher comorbidity burden, and had different absolute event rates. Additionally, OSCAR SCA cases included survivors (5.1% of total SCA cases), whereas FHS included only fatal SCD. Despite these differences, the consistency of the primary finding of higher risk of SCA after a recurrent cardiovascular event across cohorts reinforces the robustness of the findings. We reported consistent findings in the OSCAR HF cohort stratified by LVEF, but acknowledge that use of baseline LVEF rather than dynamic LVEF parameters was a limitation.

In the OSCAR, recurrent cardiovascular events may have been missed if they occurred at institutions outside of Cedars‐Sinai Health System, and cases of SCA could have been missed if they occurred outside Los Angeles County. In the FHS, the cohorts are composed primarily of individuals of European ancestry residing in New England, which may limit generalizability of the findings to other racial, ethnic, and geographic populations. For both OSCAR and FHS, the possibility of residual confounding cannot be excluded, and as both studies are observational, associations can be described but causality cannot be established.

There is a potential risk of immortal time bias as the exposure of interest occurs during follow‐up. To mitigate this risk, we used survival analyses with time‐dependent covariates, ensuring that person‐time was appropriately classified according to exposure status.

CONCLUSIONS

In a contemporary observational, prospective health system cohort study of patients with a first‐time ACS or HF hospitalization, recurrent cardiovascular events were associated with a significantly increased risk of SCA. The findings for recurrent ACS were externally validated in the community‐based FHS. These findings highlight the importance of further investigation to evaluate whether incorporating recurrent cardiovascular events into dynamic risk assessment frameworks could help to inform targeted management and prevention strategies for SCA in diverse clinical settings.

Sources of Funding

Dr Pope is a visiting postdoctoral fellow at Cedars‐Sinai Medical Center from Oslo University Hospital, Norway. The FHS (Framingham Heart Study) is funded by 75N92025D00012, 75N92019D00031, HHSN268201500001I, and N01‐HC 25195. Dr Benjamin is partially funded by the National Institutes of Health, National Heart, Lung, and Blood Institute R01HL092577.

Disclosures

None.

Supporting information

Tables S1–S6

STROBE checklist